Brain as metaphor for Zettelkasten?

by Alex Nelson, 5 July 2026

tl;dr Linking slips seems to have some relation to the structure of memory in the brain. Just as memories are “spread out” in the brain (as opposed to “localized” in some particular cluster of neurons), it’s OK for slips to be “spread out”, too. In fact, creative insights are generated this way. Some examples are given.

And, no, I didn’t use AI to write this. As any TeX user will tell you, we Mathematicians just use em-dashes a lot.

What did Luhmann say?

Luhmann writes on slip 9/8b (translation is mine, as mediocre as it is):

For the general structure of memories, see Ashby [Ashby, W. Ross (1967): The Place of the Brain in the Natural World, in: Currents in Modern Biology 1, 95-104] 1967, p. 103.

It is therefore important that one is not dependent on a numerous point-by-point [Ger.: Punkt für Punkt, more like “list of bullet points”] accesses, but rather on relations between notes, i.e., on references, which make more available at once than one might expect during an exploration [lit., “search impulse”, Ger.: Suchimpuls, Suchen for “search”, impuls for momentum, impulse, impetus, stimulus, incentive] or even a fixation of thought [Ger.: Gedanken-fixierung, presumably meaning “looking up a particular thing”?].

The relevant passage from Ashby’s article speaks about “memory traces” and Karl Lashley’s work showing that a memory is “dispersed” in the brain (as opposed to being localized to some specific neuron or cluster of neurons).

What’s a “memory trace”?

The notion of “memory trace” seems to be a synonym for “engram”. Philosophers seem to speak of memory traces, neuropsychologists speak of engrams. The best review I could find (which I could appreciate and understand) is Felipe De Brigard’s “The Nature of Memory Traces” (Google Scholar).

There has been a fairly “standard” model of “memory traces”, which has been patched over time to address inadequacieis and shortcomings.

The current model for memory formation and recall is the “multiple trace theory” (page 410 of de Brigard’s paper). Basically a bunch of neurons scattered throughout the brain store very small fragments of the same memory, and one “hippocampal index” neuron binds them all together into a single “memory”.

As I understand it, the “same” neurons can be “reused” — for example, if you have multiple memories in the same geographical location or place (for example: the same coffeeshop I proposed to my wife is the same coffeeshop I worked my first job), then the neurons associated with remembering that location (“the coffeeshop” itself) could be “reused” as you form new memories in the location.

(Thus far, the theory seems relevant for addressing the issue of “where do I place a note?”—like neurons, notes can be scattered throughout the Zettelkasten—and link-formation.)

Importantly, the part which is unique to the “multiple trace theory” addressing the inadequacies of the “standard model”, the “multiple trace theory” contends that re-indexing occurs when a memory trace gets re-activated. (I only mention this because it seems the ‘keyword index’ Luhmann uses is analogous to Hippocampal index neurons.)

How did Luhmann use this?

Luhmann appears to use the model of “memory traces” in two obvious ways: (1) in the form of the ‘keyword index’ or ‘register’ (which is analogous to the Hippocampal index neurons), (2) links between the slips of cards.

Johannes Schmidt writes in Niklas Luhmann’s Card Index: The Fabrication of Serendipity (§6) pages 58–59 (footnotes removed) about specifically how this slip 9/8b relates to the keyword index in Luhmann’s Zettelkasten:

Whereas the index to the first collection was still of fairly manageable size with its 1,250 entries, the continuous updates of the index — as another part of the data base maintenance — to the second collection ultimately resulted in 3,200 entries. Contrary to the subject index of a book, the file’s keyword index makes no claim to providing a complete list of all cards in the collection that refer to a specific term. Rather, Luhmann typically listed only one to four places where the term could be found in the file, the idea being that all other relevant entries in the collection could be quickly identified via the internal system of references described above. As Luhmann noted, this concept goes back to the general structure of the brain modeled by W.R. Ashby: the capacity of the brain does not derive from a huge number of point-to-point-accesses but on the relations between the nodes (i.e. notes). Therefore, by contrast, the large number of words listed in the keyword index indicates that this list itself was at least intended to meet the standard of (thematic) completeness, i.e. complexity of the index file.

Elsewhere, Johannes Schmidt writes in Niklas Luhmann’s Card Index: Thinking Tool, Communication Partner, Publication Machine (pg 17, most footnotes removed) about how the links on slips of paper were intended to facilitate exploration and connecting ideas in surprising ways:

Luhmann himself called his system of references a “web-like system” (spinnenartiges System). This metaphor suggests interpreting it along network-theoretical lines. A key feature explaining the productivity of this filing system is its potential for enabling ‘short cuts’, i.e., the fact that a reference may lead to a completely different (both in terms of subject and location), distant region in the network (file). Luhmann himself considered this feature, which counteracted the collection’s primary system of organization, to be of crucial significance: “The references must not capture collective concepts that aggregate key aspects but must selectively lead away from the material subsumed under them”[35] so that they facilitate interpretations and contextualizations of his notes that differed from those intended when creating and initially integrating the notes in the file system.

Footnote: [35] Luhmann, Zettelkasten II, index card no. 9/8b1

How does this help?

For me, this helps with the issue of when to form links and not to worry where to place things in the Zettelkasten.

Perhaps the best way to understand how this is useful is by examples.

Example 1. I have some notes on heterodox economics in my Zettelkasten. The notion of “time” appears frequently, because Marginalist economics works with static models. So I have some slips in my Zettelkasten from reading Termini’s Logical, mechanical and historical time in economics talking about the various distinctions of time in Economics: as a “slide show” of static models (logical time), as an exogeneous parameter for models analogous to time in classical mechanics in physics (mechanical time), and as determined by the entropic arrow of time (historical time). I can link to my slip Entropy (physics) on “entropy” in physics.

At the same time, I am now trying to understand Brouwer’s Intuitionism in the foundations of Mathematics. Brouwer starts with Kantian philosophy, arguing that a priori intuition of time suffices for getting started with arithmetic. Ah, time!? Which time are we speaking of? I have a frame-of-reference for at least three distinct notions of time which are germane for this discussion. So I can link this Intuitionism slip to the Time in economics, which then links to Entropy (physics).

Following these links allows me to relate the physics of Entropy to a “thought-experiment” describing the foundations of Mathematics. That is surprising, no one has written about this, it is meaningful and fruitful.

Example 2. I am writing notes from Dan Ingalls’s Design Principles Behind Smalltalk, because Smalltalk is not “just” a programming language but an entire “system” which you “live in”. That’s weird and unique, it’s shared by only one other programming language (Lisp).

Ingalls describes Smalltalk by introducing a “toy model” of humans consisting of a physical, tangible ‘body’ and some intangible ‘mind’. When humans communicate with each other, it is through the physical channels via “explicit communication” (through spoken word, written word, etc.) along with some “implicit communication” through shared culture and experience.

What’s clever about Ingalls’s model is that we can describe a computer as consisting of a ‘body’ (its hardware) and ‘mind’ (its current state, configuration of memory, value of variables and parameters, etc.). A human can communicate to a computer via keyboard and mouse and GUI, and the computer can reply via updating the monitor and printing things out. In this way, Smalltalk fosters ‘dialogue’ between the human and the computer to enable the two to embark on some creative endeavour.

Writing notes on this for my Zettelkasten, I have Smalltalk programming language, Dialogue with computer which links to Dialogue (and I updated Dialogue to backlink to Dialogue with computer) and Communication, Computer has body and mind and so on. However, Communication is linked to my note on Autopoiesis since (as Luhmann pointed out) communication necessitates a reply in the form of communication. This allows me to relate programming in Smalltalk as grounded implicitly in my understanding of Autopoiesis. I honestly do not know if anyone has ever related “the act of programming” with “Autopoiesis” before, but it is new to me.

Example 3 (double whammy). Why do these examples matter? Because I am writing about Proof Assistants, something like an interpreter which double checks your proofs. I am trying to present a “digital Hilbert’s programme” where the proof assistant gives the user essentially Hilbert’s finitary metatheory.

So what? Well, you are supposed to “live in” the proof assistant—just as you “live in” Smalltalk. Hilbert’s finitary metatheory is heavily inspired by Brouwer’s Intuitionism (from Example 1), where the “User” is in Dialogue with other “Users” to do Mathematics. Using my Zettelkasten, I can tie all these loose ends together, relating Autopoiesis to the practice of “doing Mathematics” to modeling user-interaction with a proof assistant to…well, you get the idea.

Connecting the “practice of using a proof assistant” to Ingalls’s Dialogue with computer to Brouwer’s Intuitionism transforms Intuitionism from a solipsistic description of Mathematics (as a solo activity executed by a single Mathematician) to a communal activity (as executed by a single Mathematician and a computer, which leads to communication with other Mathematicians over the computer).

None of these elements are particular innovative on their own. But when combined together, it presents a perspective no one discusses.

Remark. These examples are actually from my Zettelkasten. As you can see, the links are not “obscure” or “un-obvious” (anti-obvious?). They are quite “vanilla”.

The surprising thing is that two vanilla links results in novel relationships. They are not unreasonable, just “hidden” from plain sight.

The value added came from following the links. These links produced quite surprising combinations.